Analyst Trust in Government Data Portals: Causal Modeling of Interpretable Dashboards

Abstract

Government data portals serve as critical infrastructures for evidence based policy making, yet the sheer volume and complexity of available data often hinder effective decision making. A major barrier to the utilization of these platforms is the lack of analyst trust in automated predictive models, which traditionally rely on correlational rather than causal relationships. This paper explores the impact of integrating causal modeling evidence into interpretable dashboards on the trust levels of public sector data analysts. Through a comprehensive mixed methods experimental design, we compare analyst interactions with standard predictive dashboards against those equipped with interactive causal graphs and counterfactual reasoning tools. The findings demonstrate that providing explicit causal evidence significantly enhances analytical trust, reduces reliance on spurious correlations, and improves the overall accuracy of policy recommendations. Furthermore, the study reveals that while causal interpretable dashboards initially introduce a higher cognitive load, this effect dissipates over time as analysts build robust mental models of the underlying data generating processes. This research contributes to the emerging field of explainable artificial intelligence in public administration by providing empirical evidence on how causal frameworks can transform government data portals from passive repositories into trusted analytical ecosystems.